Download PDF

The Physical, Social, and Cultural Determinants of Obesity:
An Empirical Study of the U.S.
Tangel Chang
Department of Agricultural and Resource Economics
University of California, Davis
[email protected]
James M. Barrett
Department of Agricultural and Resource Economics
University of California, Davis
[email protected]
Stephen A. Vosti
Center for Natural Resources Policy Analysis
Department of Agricultural and Resource Economics
University of California, Davis
[email protected]
Poster paper prepared for presentation at the International
Association of Agricultural Economists Conference, Gold Coast, Australia,
August 12-18, 2006
Copyright 2006 by Chang, TC, Barrett, JM, Vosti, SA. All rights reserved. Readers may
make verbatim copies of this document for non-commercial purposes by any means,
provided that this copyright notice appears on all such copies
1
1. Introduction
The average Body Mass Index (BMI)1 in America has increased by 15% from 1970 to
2003, from 25.4 kg/m2 to 29.1 kg/m2. The percentage of obese people increased from 16% to
40% of the population, with 64% of them being overweight. The incidence of excess weight has
increased faster for women, faster for the least educated and for those who are married.
Currently, the United States leads the world in per capita incidence of obesity (Cutler, 2003).
Obesity is significantly correlated with coronary heart disease, hypertension,
hypercholesterolemia, gall bladder disease, stroke, and type-2 diabetes (Wann, CDC, 2005).
These illnesses have a substantial impact on U.S. health care costs. In 1995, health-care costs
associated with excess weight were $51.6 billion or 5.7 % of the total U.S. health care
expenditures. Also, excess weight decreases productivity and leads to loss of work time
(Peralta-Alva, et al., 2005).
1.1 Causes of Weight Increase in the last 30 years
Many factors, such as family lifestyles, physical inactivity, psychological disturbances,
and occupational problems can potentially contribute to the development of obesity (Flegal et al.,
2002, Hedley et al.,2004). Obesity is strongly influenced by demographic and social-economic
factors (Flegal et al., 2002, Bray, 1980). Productivity improvements, in particular in agriculture
and through the reduction of trade barriers have led to decreases in commodity prices (Alston et
al., 2005). Fast-food restaurants marketed a combination of soft drinks, fries and burger at a low
price. In 1999, the total advertising expenditures for US food products was $7.3 billion, of
which $765 million was spent to advertise candy and gum, $549 million to advertise soft drinks,
1
Body Mass Index (BMI) is defined as weight in kilograms divided by the square of
height in meters. Individuals with a BMI of 25 kg/m2 to 29.9 kg/m2 are overweight, while
individuals with a BMI of 30 kg/m2 or more are obese.
2
and $330 million was spent on snacks (Story et al., 2004). Scientists from WHO suspect that
high-fructose corn syrup (HFCS) plays an essential role in obesity (Bray, 1980). The has bee a
significant rise in the consumption of HFCS and in particular, there has been a rise in the
consumption of soft drinks which use HFCS (USDA, 2000). HFCS is highly correlated with
obesity in children (Morrill, 2004).
Technology innovations produced many time and labor-saving products, including
computers, dishwashers, and televisions, which contribute to the reduced calorie expenditures
(U.S. Department of Energy). Americans currently walk and bike less than ever, while their
mode of transportation is more often driving even for short distance trips (BLS, Time Use
Survey). In 2000, more than 26% of adults reported no leisure time physical activity (Chou et
al., 2004). There has been as substantial increase in time spent watching television and playing
computer games (U.S. Department of Labor, Time Use Survey, 2006). As a consequence,
children exercise less. One theory suggests that the major reason for the increase in excess
weight in the U.S. is changes that occurred in the technology for cooking and preparing food
leading to an increase in the number of meals Americans consume (Cutler, 2003). Ownership of
microwave ovens increased from 0% of households in 1960 to over 80% today (US DOE, 2006).
The number of households that own television sets rose from a low of 10% in the 1950’s to
nearly 100% today (Nielson Media Research, 1995). Average time spent watching television has
risen from 70 minutes per day in 1960 to 181 minutes per day in 2000 (Berg et al., 2002).
Additionally, food consumption patterns have changed. Per capita consumption of both
carbohydrates and fats as well as total energy (caloric) consumption in the last 30 years of the
twentieth century have risen substantially (USDA, 2000).
3
While many studies have identified specific changes in the eating and energy expenditure
habits, and changes in incomes in the U.S. population, none so far have attempted to assess the
relative importance of each of these factors in determining the recent rise in body weight.
2. Empirical Analysis
OLS Regressions were run using three different left-hand-side variables: BMI, weight,
and waist circumference. Logit and probit regressions are run to identify factors related to the
likelihood of being above a healthy BMI; a final tobit regression identifies factors related to the
amount by which a subject exceeds a healthy BMI.
2.1 The Data
Data are taken from the National Health and Nutrition Examination Survey, 2001-2002
(NHANES 2001-2002). This survey contains data for a total of 11,039 individuals representing
all ages. Data was collected between January 2001 and December 2002. All dietary information
was based on a 24-hour recall of food consumed the day prior to the survey. Food consumption
was recorded in grams except for cholesterol which was recorded in milligrams. Water
consumption was recorded in milliliters. All activity related questions referred to activity in the
30 days prior to the date the survey was taken by each individual. For example, the variable
dailytv was phrased “Over the past 30 days, on average how many hours per day did you sit and
watch TV or videos?” Physical measurements were obtained the same day that other data was
collected. Weights were recorded in kilograms and heights were recorded in centimeters.
Descriptive statistics are presented in Table 1. The three OLS regressions were run using
the left hand-side variables, BMI, WAIST and WEIGHT. The Tobit regression was run using the
left-hand-side variable BMIHigh which is defined as 0 if BMI is less than 25 (corresponding to
BMI for the non-overweight category) and BMI - 25 for those who have BMI>25.
4
Variable
Name
Variable
Type
BMI
Weight
Waist
Bmihigh
BmiGT25
number
number
Number
0,number
0/1
Current BMI (kg/m2)
Addsalt
Age
Agecu
Agesq
Alcho
Born
BWeight
Carb
Chol
College
Dailytv
Eatrest
Fibe
Gender
Hgt
Hgtsq
HIncome
Kcal
Lessact
Lessoth
Mar
Modact
Moreact
Moreoth
Muscle
Numfoods
Phyact
Preg
Protein
Water
Sugar
Tasks
Tfat
Vigact
Vite
Walkbike
Water
White
number
number
number
number
0/1
0/1
Number
number
number
0/1
number
number
number
0/1
number
number
number
number
0/1
0/1
0/1
0/1
0/1
0/1
0/1
number
0/1
0/1
number
number
number
0/1
number
0/1
number
0/1
number
0/1
How often do you add salt to your meal each week
Description
Current Weight in (kg)
Waist circumference (cm)
0 if bmi<25 else 25-bmi
1 if bmi>25 else 0
Current age in years
Age cubed
Age squared
Have you had more than 12 alcoholic drinks in the past year
1 indicates born in U.S.
Birth Weights in ounces
Total Carbohydrates per in grams
Total Cholesterol (mg)
1 if some or more college
Daily Hours of TV, video or computer use
How often do you eat in restaurants each week
Total Fiber per day grams
1 if Male, 0 if Female
Current Height in centimeters
Height squared
Household Income
Average kilocalories consumed per day
Do you engage in less activity than people your age
Do you engage in less activity than you did 10 years ago
1 if with a partner
Have you regularly engaged in moderate physical activity in the past 30 days
Do you engage in more activity than people your age
Do you engage in more activity than you did 10 years ago
Have you regularly engaged in Muscle building activity
Average number of different types of foods per day
Have you regularly engaged in physical activity in the past 30 days
1 if pregnant
Average Protein in grams per day
Average Water in liters per day
Average Total Sugar per day in grams
Have you regularly engaged in physical tasks around house in the last 30 days
Average Total Fat per day in grams
Have you regularly engaged in vigorous physical in the past 30 days
Average Vitamin E per day in milligrams
Have you walked or biked in the last 30 days?
Water consumed (ml)
1 if White, 0 if nonwhite
Obs
Mean
Std. dev
Min
Max
8618
8618
8500
8618
8618
24.8
64.5
84.8
2.7
0.4
7
27.1
20.5
4.6
0.5
8
9.8
39.6
0
0
65.4
196.8
165.2
40.4
1
4786
8618
8618
8618
4620
8618
4263
8618
8618
7873
8569
6717
8618
8618
8618
8618
7885
8618
8618
8618
6165
8618
8618
8618
8618
8618
8618
8618
8618
8614
8618
8618
8618
8618
8618
8618
8618
8618
1.9
31
85928
1463
0.7
0.8
116
276
269
0.3
1.8
2.3
15
0.5
157
25175
7
2127
0.3
0.1
3
0.4
0.1
0.3
0.3
15
0.1
0
76
1015
26
140
0.4
79
0.3
0.2
254
0.4
0.8
23
141759
1855
0.5
0.4
22
135
233
0.4
1.8
2.3
9
0.5
23
6442
3
1000
0.4
0.3
2.3
0.5
0.2
0.4
0.4
6
0.4
0.2
42
1259
16
88
0.5
45
0.5
0.4
124
0.5
1
2
8
4
0
0
20
0
0
0
0
0
0
0
79
6241
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
9
85
614125
7225
9
1
208
1700
3052
1
7
21
128
1
199
39641
11
15594
1
1
1
1
1
1
1
46
1
1
718
59472
228
1142
1
840
1
1
2000
1
In this manner the data is censored and has positive values that correspond to the amount by
which BMI exceeds a healthy BMI for the overweight population. For the probit and logit
regressions the value BMIGT25 was used, which is a dummy variable whose value is 1 for BMI>
5
25 and 0 otherwise. Those who have a value of 1 are by this definition overweight. In addition
to the information in Table 1 the following descriptions are informative: Addsalt is the number
of times that salt was added to a meal in the previous week. Eatrest is the number of times that
the subject reported eating in a restaurant in the previous week.2
2.2 Model Specifications
OLS:
Three OLS models were specified. The three models differed based on the specification of left
hand side (dependant) variables.
(1) BMI = α1 + β1x + ε1
(2) WEIGHT = α2 + β2x + ε2
(3) WAIST = α3 + β3x + ε3
Logit:
BMIGT25 is defined as 1 for BMI>25 and 0 otherwise.
The multinomial logit model has the form:
(4) Pj = exp(βkx)/Σkexp(βkx) (j,k=0,1)
Where Pj is the probability of the jth state occurring (in this case BMIGT25=1)
Normalizing for the normal weight category (i.e.BMIGT25=0, j=0) we have
(4’) Log(P1/P0) = exp(β1x)
2
A number of variables were dropped from the analysis since they are highly correlated with other variables. In
particular, activity variables that tracked the frequency of activities were dropped. Also, carbohydrate consumption
is highly correlated with sugar and college is correlated with income.
6
As such the coefficients β1 determines the relative probability of state 1 (overweight) to
state 0 (not overweight). Positive (negative) values of β increase (decrease) the relative
probability of being overweight.
Probit:
The probit regression is specified in the same manner as the logit. The core difference
lies in the distribution of errors. In the logit model, errors are assumed to follow the standard
logistic distribution. The errors of the probit model are assumed to follow the standard normal
distribution.
Tobit:
The tobit is a censored normal distribution. Data is censored such that those who are not
overweight such that the left-hand side variable BMIHigh is defined to be 0 for BMI<25 and BMI
- 25 for subjects with BMI>25.
The right hand side variables for each of the regressions are listed in table 001. Each of
these variables can be categorized as (a) Demographic characteristics previously reported as
correlated with weight (e.g. gender, income etc). (b) Diet characteristics (e.g. sugar, water) and
(c) activity related (e.g. dailytv).
7
3. Empirical Results3
Tables 2, 3, and 4 report the results of empirical estimations
Table 2. Determinants of BMI OLS
Table 3 Determinants of Wgt OLS
Table 4
Adjusted R-squared 0.4287
Adjusted R-squared 0.7183
Adjusted R-squared 0.6218
Root Mean Squared Error 5.3202
Root Mean Squared Error 14.436
Root Mean Squared Error 12.676
F-stat (22,6079) 209.09
F-stat (19,7836) 0.0000
F-stat (21,7836) 0.0000
Coef.
tratio
P>|t|
0.641
18.3
0
Age ***
-0.009
-9.28
0
Agesq ***
(Agecu/1000) ***
0.033
4.57
0
Born ***
1.066
5.2
Chol ***
0.001
Dailytv ***
Coef.
tratio
P>|t|
Age ***
0.606
21.21
0
0
(Agecu/1000) ***
0.046
-12.8
0
4.26
0
Chol ***
0.005
4.53
0
0.898
8.06
0
Dailytv ***
7.12
0
Gender ***
-1.563
0
Eatrest
-1.44
0.15
Hgt ***
-0.855
-4.04
10.578
0.798
0.105
0
Fibe *
1.86
0.063
0.005
19.085
0
Gender ***
0.045
1.278
-3.3
0.001
Hinc **
-0.122
-2.1
0.036
Hgt **
0.244
2
0.003
Kcal ***
-0.003
-4.86
0
(Hgtsq/1000)
1.63
0.103
Lessact **
1.255
2.49
0.013
Hinc ***
-4.73
0
Lessoth ***
6.235
12.13
0
Kcal ***
0.468
0.266
0.003
-4.19
0
Moreoth ***
-3.553
-8.27
0
Lessact
0.825
1.53
0.126
Muscle ***
-1.147
-2.82
0.005
Lessoth ***
9.53
0
0
Numfoods ***
-0.151
-4.38
0
Moreact **
-2.15
0.032
-5.22
0
Protein
0.012
1.48
0.14
-8.56
0
1.394
3.72
0
Water ***
0.001
9.14
0
-5.28
0
Protein ***
0.007
1.93
0.054
Sugar **
0.007
2.24
0.025
Numfoods ***
4.886
1.766
3.681
2.152
0.151
-4.26
0
(Water/1000) ***
0.446
8.45
0
Tfat ***
0.03
3.33
0.001
Protein *
0.015
1.86
0.061
Sugar ***
0.004
2.97
0.003
White ***
-1.455
-4.04
0
Water ***
0.001
8.37
0
Tfat ***
0.011
2.87
0.004
Cons ***
47.831
8.36
0
Sugar ***
0.009
2.66
0.008
Vigact ***
0.561
3.4
0.001
Tfat *
0.017
1.9
0.057
White ***
-0.9
-5.76
0
Cons ***
23.22
4.02
0
15.224
36.32
0
Variable
Coef.
t-ratio
P>|t|
0.963
21.479
0
-0.009
-18.18
Chol ***
0.004
0
Dailytv ***
3.06
0.002
0.309
6.52
0
Eatrest
-0.048
-1.56
0.119
Hgtsq ***
Gender ***
-0.403
-2.71
0.007
Hinc ***
-0.081
-3.33
0.001
(Kcal/1000) ***
-1.01
-4.11
0
Lessact
0.342
1.53
0.125
Lessoth ***
2.224
10.38
0
Moreact **
-0.766
-2.24
0.025
Moreoth ***
-1.251
-6.96
Numfoods ***
-0.075
Preg ***
Age ***
Agesq ***
Cons ***
3
Waistst – OLS
Variable
Variable
Moreoth ***
Muscle ***
*** Significant at 0.001, **Significant at 0.05, *Significant at 0.1
8
3.1 Results of OLS multivariate regression with BMI as the dependent variable
The results of the OLS regression are displayed in Table 2. BMI increases with age. All
three components (Age, Agesq and Agecu) are significant, with the linear and cubic terms
positive and the squared term negative. The linear and cubic terms are dominant implying that
BMI increases as a function of age. The unfortunate implication of this is that with all else
equal, if you live long enough you are likely to become overweight.
BMI increases with each daily hour of television or computer use. The coefficient on
Dailytv is 0.309 implying that each additional daily hour will add 0.309 to one’s BMI (on
average). Consequently, this factor alone can make the difference between a healthy BMI (20 to
25) and a BMI considered overweight. Six hours per day spent on these activities would add 1.9
to BMI moving someone with a healthy BMI of 23.5 would into the overweight category. 15%
of the sample reported dailytv of 5 or greater. Of these, 62% have BMI greater than 25.
The coefficient on white is -0.900 implying that Caucasians on average have a lower BMI
of 0.900 compared to non-whites, ceteris paribus. Men (gender = 1) when controlling for other
all other factors on average have a BMI 0.525 less than women, ceteris paribus. Hincome has a
negative coefficient, implying high income have lower BMI. Chol, sugar, Tfat and Protein all
add to BMI consistent with expectations. Water also added to BMI. Interestingly, the sign on
Kcals (the number of kilocalories consumed in a day) was negative and highly significant
(p<0.001), implying that an increase in kilocalories decreases BMI. This result is suspect since
even the most naïve model suggests the opposite. However, when BMI is regressed against kcal
alone the sign is very slightly positive. There may be systematic under-reporting of food
consumption by those overweight, or at the time of the survey, those who were overweight
responded to the survey by decreasing their consumption, biasing the results. Numfoods has a
9
negative coefficient implying that a diverse diet leads to lower BMI. Born has a positive
coefficient indicating that those born in the U.S. are more likely to have an elevated BMI. Preg
has a positive coefficient indicating elevated BMI for pregnant women.
Moreoth and Moreact decreased BMI and Lessoth increased BMI, implying that more
activity relative to the past and to others decreases BMI and less activity relative to others
increases BMI. The coefficient on Lessact was insignificant and the coefficient on Vigact was
positive. This positive coefficient implies that vigorous activity increases BMI. This is opposite
to our expectation; however vigorous activity can lead to muscle, which is denser than fat and
consequently would elevate BMI. Unfortunately the coefficient on muscle was insignificant.
The coefficient on Eatrest was insignificant; as reported in the survey data eating in
restaurants does not in itself lead to excess weight. This is contrary to our expectations.
3.2 Results of the OLS with weight and waist as the dependent variables
The results of the regressions for weight and waist are displayed in Tables 3 and 4.
Results of the regressions with weight or waist as the dependent variable were consistent with
the OLS regression on BMI except for the following:
Height and Hgtsq were added as independent variables. For both the Weight and Waist
the impact of Height was positive as expected. The coefficient on Born was insignificant in
both cases indicating that being born in the US did not have the same impact on Weight and
Waist as it does on BMI. This may indicate that those born in the U.S. have higher BMI but not
physical stature. Lessact was significant and negative for weight indicating that more activity
than the past leads to less weight. Muscle decreases both weight and waist indicating that muscle
building diminishes weight and waist circumference. Preg and Vigact were not significant for
either regression.
10
3.3 Results for tobit regression
The tobit regression identifies factors that influence the degree of excess weight among
those already defined as overweight (Bmihigh>0); 3874 subjects out of our sample of 8618 have
Bmihigh>0. The results of this regression are displayed in Table 5.
Age contributes to an increase in excess weight. Each hour of television watching or
computer use (Dailytv) adds 0.388 points to Bmihigh. This result is higher than the coefficient on
dailytv (0.309) from the regression with BMI implying that the impact on BMI is more
substantial for the population segment with elevated BMI. Muscle-building activity (Muscle)
decreases Bmhigi. However, the results from the OLS regression of BMI on muscle-building
activity were insignificant, implying that the impact of these activities, for those in the
population who are not overweight is ambiguous, whereas for those who are overweight, the
result is beneficial. Lessoth, Moreoth and Moreact all have coefficients larger than those in the
OLS regression, indicating that activity and exercise (or lack of exercise) has a more pronounced
impact on those who are overweight. The results for Gender and White were also greater for the
tobit regression, suggesting that women and nonwhites, once overweight, are likely to be even
more overweight than men and whites.
11
Table 5: Tobit Regression BMIHigh
Table 6: Logit regression – BmiGT25
Table 7 Probit regression - BmiGT25
Number of obs =
Number of obs = 8565
Number of obs =
LR chi2(18)
=
2877.29
7836
LR chi2(20)
= 3175
LR chi2(17)
=
3158.42
Prob > chi2
=
0.0000
Prob > chi2
=
Prob > chi2
=
0.0000
0.0000
8565
Log likelihood = -13763.242
Log likelihood = -4306.2619
Log likelihood = -4314.85
Pseudo R2
Pseudo R2
Pseudo R2
=
Var
0.0946
Coef.
T
(Agecu/1000) ***
0.073
6.7
0
Age ***
0.976
16.64
10.38
0
Age ***
0
Agesq ***
Agesq ***
-0.015
P>|t|
=
Var
0.2694
Coef.
0.116
t
P>|t|
=
Var
0
Age ***
-0.001
16.56
13.39
0
Agesq ***
Born *
0.131
1.75
0.08
Born *
Coef.
0.071
t
P>|t|
-0.001
17
13.74
0
0
0.075
1.68
0.094
0.25
3.09
0.002
Chol ***
0.002
4.27
0
(Chol/1000)***
0.417
3.04
0.002
Dailytv ***
0.388
6.36
0
Dailytv ***
0.059
3.63
0
Dailytv ***
0.036
3.72
0
Gender ***
-0.651
-3.12
0.002
Hgt ***
0.226
7.02
0
Hgt ***
0.101
6.16
0
(Hgtsq/1000) ***
-0.641
-6.46
0
Hgtsq/1000 ***
0.282
-5.49
0
(Kcal/1000) ***
-0.199
-3.62
0
Kcal/1000***
-0.125
-3.78
0
Lessact **
0.181
2.41
0.016
Lessact **
0.108
2.42
0.016
Lessoth ***
0.374
4.8
0
Lessoth ***
0.225
5.7
0
Moreoth ***
-0.37
-5.78
0
Moreoth ***
-0.219
-5.7
0
Muscle ***
-0.269
-4.19
0
Muscle ***
-0.158
-4.08
0
Numfoods ***
-0.023
-4.35
0
Numfoods ***
-0.013
-4.17
0
Preg ***
0.702
5.38
0
Preg ***
0.446
5.69
0
(Water/1000) ***
0.139
5.77
0
Water/1000 ***
0.066
5.96
0
Tasks *
0.118
1.91
0.056
Tasks **
0.078
2.09
0.036
Tfat *
0.002
1.73
0.083
Tfat *
0.001
1.84
0.066
-0.134
-2.24
0.025
Walkbike **
-0.08
-2.21
0.027
2.24
6.132
0.025
-7.85
0
Hinc ***
-0.078
-2.25
0.025
Kcal***
-0.001
-4.42
0
Lessoth ***
2.804
10.42
0
Moreact **
-1.047
-2.64
0.008
Moreoth ***
-1.741
-7.36
0
Muscle **
-0.581
-2.43
0.015
Numfoods ***
-0.104
-5.35
0
Sugar **
0.003
1.8
0.073
Water ***
0.001
7.77
0
Tfat ***
0.018
3.62
0
Walkbike **
Vigact *
White ***
Cons ***
Chol/1000 ***
0.2679
0.593
2.55
0.011
Vigact **
0.138
2.13
0.033
Vigact **
0.087
-0.929
13.919
-4.43
20.92
0
White ***
-6.18
0
White ***
0
Cons ***
-0.379
21.781
-8.41
0
Cons ***
-0.221
10.243
12
0
3.4 Results of logit and probit
The major difference between the logit and probit regressions and the OLS is the
interpretation of the coefficients. A factors with positive (negative) coefficients for probit or
logit increase (decrease) the probability of high BMI. The results from these regressions are
displayed in Tables 6 and 7. These results are almost completely consistent with the OLS and
tobit regressions. The following are the differences:
The coefficient on Walkbike is negative suggesting that walking or biking decreases the
probability of having BMI>25. The coefficient on Tasks is positive suggesting that doing house
tasks increases the probability of having BMI>25. This is inconsistent with expectations
although it may indicate housewives, which is consistent with expectations. The correlation
between tasks and married females is high.
In general, eating more food leads to a greater tendency for high BMI. Being a woman, a
non-white or poor will lead to an increase in BMI. Activity is generally beneficial, in particular
activity and muscle building will benefit those who are overweight and decrease the probability
of becoming overweight. Eating a diverse diet is beneficial and will diminish excess BMI and
decrease the probability of becoming overweight. Television or computer use will increase the
probability excess weight and increase BMI overall and is particularly detrimental to those who
are already overweight.
Other than Kcal and Vigact having the opposite sign expected and Eatrest not being
significant, there were no real surprises in the results. Certain demographic characteristics,
including gender, ethnic background and income are associated with weight. Those who eat
more and engage in less active lifestyles are more likely to be overweight.
13
4. Discussion
This study used nationally representative survey data to identify the effects of selected
personal characteristics and habits on obesity. Several important results emerged, some of which
are relevant for policy.
Since most Americans can currently have incomes sufficient to afford overeating and
food as a share of overall all budget is small, price changes through taxes and subsidies alone
may be insufficient to to decrease the incidence of obesity. Additional efforts aimed at
behavior modification will also be needed. Our results suggest that activity levels in addition to
food intake is important. As such, activity levels will need to be addressed. Some behavior
modifications will be relatively easy (e.g., hours spent watching television appears to have
peaked) while others will be more challenging (e.g., hours spent before computers and electronic
games are on the rise, especially among the young).
Our results add to the research that demonstrates that some groups are much more at risk
at becoming overweight and obese than others. In particular, we should pay particular attention
to women, children, non-whites, those who are not college educated, and the poor.
These results suggest a number of areas to emphasize in any policy. Television watching
and other sedentary activities such as video games are problematic. Also, considering the
positive impacts of exercise, policies directed towards encouraging active lifestyles will be
beneficial. Given the propensity of excess weight amongst certain groups (e.g. women, children
etc.), programs that encourage exercise and activity in place of television and computer games
for these individuals will be particularly beneficial.
14
Certainly exercise programs directed at those who are currently obese will be
advantageous. The results on muscle building activities for those overweight, suggest that this
type of exercise can be particularly valuable.
Diet is obviously important; however the surprising results on total calories consumed
suggest that it may be more relevant to pay attention to how much of each type of food is being
consumed rather than the overall level of consumption. In particular, sugar and fat are areas of
concern. Particularly interesting is the result on the number of foods consumed in a day. The
positive coefficient on this variable suggests that a diverse diet in itself is a good practice.
Encouraging the young, women and other groups with a high incidence to obesity to increase the
number of foods consumed will be helpful.
Finally, the results on age show that the older we get the higher the incidence of excess
weight. This suggests that older folks should pay attention to their diet and exercise. Programs
directed towards older segments of the population may produce positive results. In general, the
results suggest that a specific programs directed toward specific segments of the population are
likely to be constructive.
15
References
Alston , JM., Sumner, DA, and Vosti, SA “The Effect of Agricultural Research and Farm
Subsidy Policies on Human Nutrition and Obesity”, Selected Paper prepared for the AAEA
annual meeting, Providence, Rhode Island, July 24-27, 2005
Berg, K, Ridder, EM (Hg.), 2002: Massenkommunikation VI (Schriftenreihe Media
Perspektiven, Bd. 16).
Bray, GA, Ed. Obesity in America. New York: U.S. Government, 1980
(CDC) U.S. Department of Health and Human Services Centers for Disease Control and
Prevention National Center for Health Statistics, 2005, www.cdc.gov/nchs
Chou, SY, Grossman, M, and Saffer, H , “An Economic Analysis of Adult Obesity: Results from
the Behavioral Risk Factor Surveillance System”, Journal-of-Health-Economics. May
2004; 23(3): 565-87
Cutler, D. Glaeser, EL and Shapiro, JM, “Why Have Americans Become More Obese?” Journal
of Economic Perspectives 17, 93-118, 2003
Flegal KM, Carroll MD, Ogden CL, Johnson CL. “Prevalence and trends in obesity among US
adults, 1999-2000”, JAMA 288:1723-7. 2002
Hedley, AA, Ogden, CL, Johnson, CL, Carroll, MD, Curtin, LR, Flegal, KM, “Overweight and
obesity among US children, adolescents, and adults, 1999-2002” JAMA
2004:291:2847.2850
Morrill, AC, Chinn, CD. “The obesity epidemic in the United States.”
J Public Health Policy. 2004;25(3-4):353-66.
Nielson Media Research, 1995
Peralta AA & Pere Gomis, P, "The macroeconomics of obesity in the United States,"
Macroeconomics 0503014, Economics WPA, revised 28 Mar 2005
Story, M and French, S, “Food Advertising and Marketing Directed at Children and Adolescents
in the US, Int J Behav Nutr Phys Act., 2004; 1:3.
USDA, 2000 “Major Trends in the U.S. Food Supply 1909-1999” Food Review 23:1, 2000
(http://www.ers.usda.gov/publications/foodreview/jan2000/)
U.S. Department of Energy http://www.energy.gov/
U.S. Department of Labor Statistics http://www.bls.gov
16